PlatformMinis
One job, said once. Done every time.
A Mini is a small agent with a single standing job. Say what it should do in your own words and it sets itself up: when it wakes, where it reports and what it may do. By default it changes nothing without you.
Say it once. It keeps happening.
Minis live in a section of their own, and on the rail at chat.salexhub.com, beside your conversations.
Describe it, and it sets itself up
Write the job the way you would say it. The Agent turns it into a Mini and shows you how it understood you, when it will wake and that it changes nothing on its own, before you create it. A ready Mini needs no model and no AI credits to set up.
When it wakes
On the clock: every few hours on the quarter hour, on the days you choose at a time, or on a day of the month. On an event, such as a deal reaching Won. On a watch over a saved view, or over records nothing has happened on for a number of days. When a goal changes pace. Or when you run it now.
Where it reports
In the app, by email, or both, and never nowhere: a Mini that runs and tells no one is the same as one that does not run. A burst of events can be gathered into one run and one report instead of ten.
A thread of its own
Each run opens with when and why it woke, then the report. Reply in that thread to change the job: make it shorter, only on weekdays, skip the closed deals. The Mini changes with your reply.
What it may do
It reads and reports, or it proposes: every change it wants arrives as a card you approve in its thread, and runs as you when you say yes. A Mini your own Agent holds never acts by itself. One held by an agent you hired acts inside that agent's limits, and anything that needs approval still waits.
Every run on the record
Its history keeps each run: what woke it, how it ended, what it cost in AI credits, whether the report reached you, and anything it was refused. Pause it, edit it, archive it, bring it back.
Your connected services
A Mini can use the MCP servers an admin connected to your organization. The tools an admin marked as reads run on their own; anything else asks first.
Domain Skills Coming soon
Sales and marketing method the platform carries for you: which qualification fits this deal, what is already known, what is missing and what to propose. A Skill applies itself when the facts call for it, and what it finds reaches the record as a proposal you approve.
The Agent makes them and reads them.
Asking for a Mini in a conversation is the same as making one on its page: the draft comes back for you to look at, and creating it waits for your approval unless you work in Full autonomy.
minis.Turns what you wrote into a Mini: a name, its instructions, its wake and its delivery.Asks by defaultdraft minis.Makes the Mini, held by your own Agent or by an agent you hired.Asks unless Full autonomycreate minis.Changes its name, instructions, wake, delivery, mode or connected services.Asks unless Full autonomyupdate minis.Lists your Minis, and the Minis of hired agents for whoever administers them.Runslist minis.Reads the last runs: what woke each one, how it ended and what it spent.Runshistory minis.Runs a Mini once, now, outside its wake.Asks unless Full autonomyrun_ now minis.Stops it waking until it is resumed; a run already going finishes.Asks by defaultpause
Each square is a mode, left to right: Ask every time (the default), Auto inside the organization, Auto including replies and Full autonomy. A filled square means the Agent runs the tool in that mode without asking.
It stands on the rest of the platform.
- The AgentA Mini is the same Agent with one standing job, running under the rights of whoever holds it.
- WorkflowsA Mini's event wake is a workflow's trigger: the same object, matched by the same rule.
- Reports & dashboardsA Mini can wake when a goal's pace changes, and its report opens with the number as the platform measured it.
- Objects & dataA watch follows a saved view of any object, your own included, or the records nothing has touched for a while.
- Control & auditIt never sees more than its owner, and approval rules hold what it proposes as they hold anyone's act.
What a Mini may never do.
- A Mini runs with the rights of whoever holds it, and never above them.
- A Mini held by your own Agent reports or proposes; it never acts unattended, because an act nobody decided is not yours.
- Every run is an AI task with its own cost in AI credits, kept in the Mini's history.
- A watch is checked every hour over at most 1,000 records, and its first check only notes what is already there, so it wakes on a change and not on the past.
- One person keeps up to 50 live Minis.
- A paused Mini wakes again from its next slot: missed slots are not made up.
What shipped lately.
All changesMinis use your connected services
A Mini can read from the MCP servers an admin connected: the tools marked as reads run on their own, and anything else asks first.
Minis
A Mini does one job that repeats: it wakes on a schedule, on an event, when a record enters or leaves a dynamic list, or when the pace of a goal changes, and its report arrives in the app or by email.
What a Mini may do, and ready ones to start from
A Mini reads and reports, or proposes: every change it wants becomes a card you approve in its thread, and only a Mini held by an agent you hired acts by itself, inside that agent's limits. A ready Mini is made without a model and without AI credits.
The Agent can read the public web
The Agent opens a public page at an address you or a search result gave it, and treats what it reads there as information, never as instructions.
Read more.
- DocsAutonomy and approvalsWhat the Agent may do on its own, and what waits for a person whatever the setting.
- DocsConnect an AI providerWhere the AI credits a run spends come from, and how to bring your own key.
- DocsWork from the chatWork with the Agent at chat.salexhub.com, the second way into the same organization.
- BlogWhy the Agent asks firstWhy a change waits for a person by default, even when nobody is watching.
One platform for the whole business.
Every module works on the same records, and the Agent works on them with your permissions.